pytorch-backbone-benchmark

Benchmarks for popular neural network models supported by timm

• CPU: Intel(R) Core(TM) i5-10400 CPU @ 2.90GHz 2.90 GHz
• GPU: RTX 3070 8GB
• RAM: 32.0 GB
• Pytorch Version: 1.8.1
• Input Tensor Shape: 1x3x608x608

All timing experiments are averaged over 100 times.

sorted by execution time

The graph is here.

	model	top1	param_count	execution time(sec)	fps
0	swsl_resnet18	73.276	11.69	0.0044	227
1	resnet18	69.748	11.69	0.0045	222
2	gluon_resnet18_v1b	70.836	11.69	0.0045	222
3	ssl_resnet18	72.61	11.69	0.0045	222
4	resnet18d	72.26	11.71	0.0048	208
5	tf_mobilenetv3_small_minimal_100	62.906	2.04	0.0055	181
6	legacy_seresnet18	71.742	11.78	0.0065	153
7	tf_mobilenetv3_large_minimal_100	72.248	3.92	0.0073	136
8	efficientnet_lite0	75.484	4.65	0.0074	135
9	ese_vovnet19b_dw	76.798	6.54	0.0074	135
10	gluon_resnet34_v1b	74.588	21.8	0.0074	135
11	tv_resnet34	73.312	21.8	0.0075	133
12	resnet34	75.11	21.8	0.0076	131
13	mobilenetv2_140	76.516	6.11	0.0076	131
14	regnetx_002	68.762	2.68	0.0076	131
15	dla34	74.63	15.74	0.0077	129
16	tf_efficientnet_lite0	74.83	4.65	0.0077	129
17	mnasnet_100	74.658	4.38	0.0078	128
18	efficientnet_es_pruned	75	5.44	0.0079	126
19	efficientnet_es	78.066	5.44	0.0079	126
20	mobilenetv2_100	72.97	3.5	0.008	125
21	tf_efficientnet_es	76.594	5.44	0.008	125
22	resnet26	75.292	16	0.008	125
23	dla46x_c	65.97	1.07	0.0083	120
24	resnet34d	77.116	21.82	0.0083	120
25	dla46_c	64.866	1.3	0.0084	119
26	gernet_s	76.916	8.17	0.0084	119
27	resnest14d	75.506	10.61	0.0084	119
28	hardcorenas_a	75.916	5.26	0.0086	116
29	selecsls42b	77.174	32.46	0.0086	116
30	resnet26d	76.696	16.01	0.0087	114
31	gernet_m	80.732	21.14	0.0088	113
32	semnasnet_100	75.448	3.89	0.0093	107
33	ecaresnet26t	79.854	16.01	0.0094	106
34	regnetx_006	73.852	6.2	0.0094	106
35	skresnet18	73.038	11.96	0.0094	106
36	spnasnet_100	74.084	4.42	0.0095	105
37	fbnetc_100	75.124	5.57	0.0096	104
38	mobilenetv2_110d	75.036	4.52	0.0096	104
39	tf_efficientnet_lite1	76.642	5.42	0.0097	103
40	regnetx_008	75.038	7.26	0.0097	103
41	tf_efficientnet_lite2	77.468	6.09	0.0098	102
42	tf_mobilenetv3_small_100	67.922	2.54	0.0105	95
43	hardcorenas_b	76.538	5.18	0.0106	94
44	efficientnet_em	79.252	6.9	0.0106	94
45	gernet_l	81.354	31.08	0.0107	93
46	dla60x_c	67.892	1.32	0.0108	92
47	cspresnet50	79.574	21.62	0.0108	92
48	tf_efficientnet_em	78.13	6.9	0.0108	92
49	selecsls60	77.982	30.67	0.011	90
50	tf_mobilenetv3_small_075	65.716	2.04	0.011	90
51	tf_efficientnet_lite3	79.82	8.2	0.0111	90
52	regnety_002	70.252	3.16	0.0112	89
53	repvgg_a2	76.46	28.21	0.0112	89
54	regnetx_016	76.95	9.19	0.0113	88
55	efficientnet_b0	77.698	5.29	0.0114	87
56	selecsls60b	78.412	32.77	0.0114	87
57	mobilenetv3_large_100	75.766	5.48	0.0115	86
58	mobilenetv3_large_100_miil	77.916	5.48	0.0115	86
59	hardcorenas_c	77.054	5.52	0.0115	86
60	mobilenetv2_120d	77.284	5.83	0.0116	86
61	legacy_seresnext26_32x4d	77.104	16.79	0.0116	86
62	tf_efficientnet_b0_ns	78.658	5.29	0.0116	86
63	tf_mobilenetv3_large_100	75.518	5.48	0.0116	86
64	resnest26d	78.478	17.07	0.0117	85
65	tf_efficientnet_b0	76.848	5.29	0.012	83
66	legacy_seresnet34	74.808	21.96	0.012	83
67	tf_efficientnet_b0_ap	77.086	5.29	0.012	83
68	mobilenetv3_rw	75.634	5.48	0.0122	81
69	seresnext26d_32x4d	77.602	16.81	0.0123	81
70	seresnext26t_32x4d	77.986	16.81	0.0123	81
71	regnetx_004	72.396	5.16	0.0124	80
72	ese_vovnet39b	79.32	24.57	0.0126	79
73	ecaresnet50d_pruned	79.716	19.94	0.0127	78
74	rexnet_100	77.858	4.8	0.0127	78
75	ecaresnetlight	80.462	30.16	0.0127	78
76	rexnet_130	79.5	7.56	0.0127	78
77	tv_resnet50	76.138	25.56	0.0127	78
78	swsl_resnet50	81.166	25.56	0.0127	78
79	tf_mobilenetv3_large_075	73.438	3.99	0.0128	78
80	repvgg_b0	75.152	15.82	0.0128	78
81	resnet50	79.038	25.56	0.0128	78
82	ssl_resnet50	79.222	25.56	0.0128	78
83	rexnet_150	80.31	9.73	0.0128	78
84	regnety_008	76.316	6.26	0.0128	78
85	gluon_resnet50_v1b	77.58	25.56	0.0129	77
86	dla60	77.032	22.04	0.013	76
87	rexnet_200	81.632	16.37	0.0131	76
88	gluon_resnet50_v1c	78.012	25.58	0.0133	75
89	regnety_006	75.246	6.06	0.0133	75
90	gluon_resnet50_v1d	79.074	25.58	0.0135	74
91	resnet50d	80.53	25.58	0.0136	73
92	regnety_004	74.034	4.34	0.0137	72
93	tf_efficientnet_lite4	81.536	13.01	0.0139	71
94	cspdarknet53	80.058	27.64	0.014	71
95	cspresnext50	80.04	20.57	0.0141	70
96	hardcorenas_e	77.794	8.07	0.0143	69
97	resnetblur50	79.286	25.56	0.0143	69
98	tf_efficientnet_cc_b0_4e	77.306	13.31	0.0144	69
99	tf_efficientnet_cc_b0_8e	77.908	24.01	0.0145	68
100	res2net50_48w_2s	77.522	25.29	0.0145	68
101	legacy_seresnet50	77.63	28.09	0.0147	68
102	regnetx_032	78.172	15.3	0.0147	68
103	hardcorenas_f	78.104	8.2	0.0147	68
104	ecaresnet50t	82.346	25.57	0.0147	68
105	ecaresnet50d	80.592	25.58	0.0148	67
106	mixnet_s	75.992	4.13	0.0148	67
107	gluon_resnet50_v1s	78.712	25.68	0.015	66
108	ghostnet_100	73.978	5.18	0.0151	66
109	seresnet50	80.274	28.09	0.0152	65
110	efficientnet_el_pruned	80.3	10.59	0.0153	65
111	efficientnet_el	81.316	10.59	0.0154	64
112	tf_efficientnet_el	80.25	10.59	0.0154	64
113	dpn68	76.318	12.61	0.0156	64
114	hardcorenas_d	77.432	7.5	0.0156	64
115	efficientnet_b1	78.794	7.79	0.0157	63
116	tf_mixnet_s	75.65	4.13	0.0158	63
117	tf_inception_v3	77.856	23.83	0.0159	62
118	inception_v3	77.438	23.83	0.016	62
119	adv_inception_v3	77.582	23.83	0.016	62
120	xception	79.052	22.86	0.016	62
121	regnetx_040	78.482	22.12	0.016	62
122	hrnet_w18_small	72.342	13.19	0.016	62
123	gluon_inception_v3	78.806	23.83	0.0161	62
124	efficientnet_b2	80.612	9.11	0.0161	62
125	tf_efficientnet_b2_ap	80.3	9.11	0.0161	62
126	efficientnet_b1_pruned	78.236	6.33	0.0161	62
127	efficientnet_b2_pruned	79.916	8.31	0.0162	61
128	dla60x	78.246	17.35	0.0162	61
129	resnetv2_50x1_bitm	80.172	25.55	0.0163	61
130	tf_efficientnet_b1	78.826	7.79	0.0163	61
131	tf_efficientnet_b1_ns	81.388	7.79	0.0164	60
132	resnetrs50	79.892	35.69	0.0164	60
133	tf_efficientnet_b1_ap	79.28	7.79	0.0164	60
134	tf_efficientnet_b2_ns	82.38	9.11	0.0165	60
135	tf_efficientnet_b2	80.086	9.11	0.0165	60
136	res2net50_26w_4s	77.964	25.7	0.0165	60
137	tv_resnext50_32x4d	77.62	25.03	0.0167	59
138	ssl_resnext50_32x4d	80.318	25.03	0.0167	59
139	dpn68b	79.216	12.61	0.0167	59
140	resnest50d_1s4x24d	80.988	25.68	0.0168	59
141	swsl_resnext50_32x4d	82.182	25.03	0.0168	59
142	gluon_resnext50_32x4d	79.354	25.03	0.0168	59
143	resnext50_32x4d	79.768	25.03	0.0168	59
144	res2next50	78.246	24.67	0.0172	58
145	nf_resnet50	80.694	25.56	0.0173	57
146	skresnet34	76.912	22.28	0.0174	57
147	resnext50d_32x4d	79.676	25.05	0.0174	57
148	repvgg_b1g4	77.594	39.97	0.0175	57
149	mixnet_l	78.976	7.33	0.0177	56
150	dla60_res2net	78.464	20.85	0.0178	56
151	vgg11	69.024	132.86	0.0178	56
152	mixnet_m	77.26	5.01	0.0179	55
153	resnest50d	80.974	27.48	0.018	55
154	tf_efficientnet_b3_ns	84.048	12.23	0.0182	54
155	efficientnet_b3	82.242	12.23	0.0185	54
156	tf_efficientnet_b3	81.636	12.23	0.0186	53
157	vgg11_bn	70.36	132.87	0.0187	53
158	xception41	78.516	26.97	0.0188	53
159	tf_efficientnet_b3_ap	81.822	12.23	0.0189	52
160	efficientnet_b3_pruned	80.858	9.86	0.0189	52
161	tf_mixnet_m	76.942	5.01	0.019	52
162	tf_mixnet_l	78.774	7.33	0.0191	52
163	eca_nfnet_l0	82.588	24.14	0.0191	52
164	dla60_res2next	78.44	17.03	0.0192	52
165	gluon_seresnext50_32x4d	79.918	27.56	0.0194	51
166	seresnext50_32x4d	81.266	27.56	0.0194	51
167	legacy_seresnext50_32x4d	79.078	27.56	0.0195	51
168	regnety_032	82.724	19.44	0.0196	51
169	regnety_040	79.22	20.65	0.0199	50
170	regnetx_064	79.072	26.21	0.02	50
171	dla102	78.032	33.27	0.0201	49
172	regnetx_080	79.194	39.57	0.0203	49
173	tf_efficientnet_cc_b1_8e	79.308	39.72	0.0204	49
174	nfnet_l0	82.76	35.07	0.0205	48
175	resnest50d_4s2x40d	81.108	30.42	0.0214	46
176	tv_resnet101	77.374	44.55	0.0219	45
177	tv_densenet121	74.738	7.98	0.0219	45
178	densenet121	75.578	7.98	0.022	45
179	res2net50_26w_6s	78.57	37.05	0.0221	45
180	densenetblur121d	76.588	8	0.0221	45
181	gluon_resnet101_v1b	79.306	44.55	0.0221	45
182	mixnet_xl	80.476	11.9	0.0224	44
183	gluon_resnet101_v1c	79.534	44.57	0.0226	44
184	vgg13	69.926	133.05	0.0227	44
185	gluon_resnet101_v1d	80.414	44.57	0.0228	43
186	resnet101d	83.022	44.57	0.0228	43
187	efficientnet_b4	83.428	19.34	0.023	43
188	tf_efficientnet_b4_ns	85.162	19.34	0.0232	43
189	wide_resnet50_2	81.456	68.88	0.0232	43
190	tf_efficientnet_b4	83.022	19.34	0.0232	43
191	repvgg_b1	78.366	57.42	0.0233	42
192	tf_efficientnet_b4_ap	83.248	19.34	0.0235	42
193	regnety_016	77.862	11.2	0.0238	42
194	regnety_064	79.722	30.58	0.0239	41
195	skresnext50_32x4d	80.156	27.48	0.0243	41
196	repvgg_b2g4	79.366	61.76	0.0243	41
197	ecaresnet101d_pruned	80.818	24.88	0.0244	40
198	gluon_resnet101_v1s	80.302	44.67	0.0244	40
199	nf_regnet_b1	79.306	10.22	0.0245	40
200	vgg13_bn	71.594	133.05	0.0246	40
201	regnety_080	79.876	39.18	0.0247	40
202	ecaresnet101d	82.172	44.57	0.025	40
203	dla102x	78.51	26.31	0.0252	39
204	res2net50_14w_8s	78.15	25.06	0.0263	38
205	gluon_xception65	79.716	39.92	0.0265	37
206	xception65	79.552	39.92	0.0266	37
207	inception_v4	80.168	42.68	0.027	37
208	vgg16	71.594	138.36	0.0273	36
209	legacy_seresnet101	78.382	49.33	0.0274	36
210	tf_efficientnet_b5_ns	86.088	30.39	0.0276	36
211	tf_efficientnet_b5	83.812	30.39	0.0277	36
212	tf_efficientnet_b5_ap	84.252	30.39	0.0277	36
213	res2net50_26w_8s	79.198	48.4	0.0282	35
214	resnetrs101	82.288	63.62	0.0283	35
215	ssl_resnext101_32x4d	80.924	44.18	0.0285	35
216	swsl_resnext101_32x4d	83.23	44.18	0.0285	35
217	gluon_resnext101_32x4d	80.334	44.18	0.0287	34
218	regnetx_120	79.596	46.11	0.0288	34
219	hrnet_w18_small_v2	75.114	15.6	0.029	34
220	vgg16_bn	73.35	138.37	0.0293	34
221	resnetv2_101x1_bitm	82.212	44.54	0.0294	34
222	densenet169	75.906	14.15	0.0302	33
223	res2net101_26w_4s	79.198	45.21	0.0305	32
224	densenet161	77.358	28.68	0.0312	32
225	repvgg_b3g4	80.212	83.83	0.0313	31
226	regnety_120	80.366	51.82	0.0314	31
227	dla169	78.688	53.39	0.0315	31
228	tv_resnet152	78.312	60.19	0.0318	31
229	vgg19	72.368	143.67	0.0321	31
230	gluon_resnet152_v1b	79.686	60.19	0.0322	31
231	dpn92	80.008	37.67	0.0324	30
232	tf_efficientnet_b6	84.11	43.04	0.0325	30
233	tf_efficientnet_b6_ns	86.452	43.04	0.0325	30
234	gluon_resnet152_v1c	79.91	60.21	0.0326	30
235	tf_efficientnet_b6_ap	84.788	43.04	0.0326	30
236	resnet152d	83.68	60.21	0.0327	30
237	gluon_resnet152_v1d	80.474	60.21	0.0327	30
238	repvgg_b2	78.792	89.02	0.0332	30
239	gluon_seresnext101_32x4d	80.904	48.96	0.0333	30
240	dm_nfnet_f0	83.342	71.49	0.0334	29
241	legacy_seresnext101_32x4d	80.228	48.96	0.0337	29
242	xception71	79.874	42.34	0.034	29
243	gluon_resnet152_v1s	81.016	60.32	0.0341	29
244	vgg19_bn	74.214	143.68	0.0342	29
245	resnest101e	82.89	48.28	0.0342	29
246	eca_nfnet_l1	84.008	41.41	0.0352	28
247	densenet201	77.286	20.01	0.0361	27
248	dla102x2	79.448	41.28	0.0364	27
249	regnetx_160	79.856	54.28	0.0365	27
250	regnety_160	83.686	83.59	0.0396	25
251	legacy_seresnet152	78.66	66.82	0.0408	24
252	resnetrs152	83.712	86.62	0.041	24
253	seresnet152d	84.362	66.84	0.0413	24
254	inception_resnet_v2	80.458	55.84	0.0426	23
255	ens_adv_inception_resnet_v2	79.982	55.84	0.0427	23
256	resnet200d	83.962	64.69	0.0429	23
257	dpn98	79.642	61.57	0.0435	22
258	tf_efficientnet_b7	84.936	66.35	0.0441	22
259	tf_efficientnet_b7_ap	85.12	66.35	0.0441	22
260	tf_efficientnet_b7_ns	86.84	66.35	0.0441	22
261	wide_resnet101_2	78.856	126.89	0.0441	22
262	repvgg_b3	80.492	123.09	0.0443	22
263	swsl_resnext101_32x8d	84.284	88.79	0.0453	22
264	ig_resnext101_32x8d	82.688	88.79	0.0454	22
265	resnext101_32x8d	79.308	88.79	0.0454	22
266	ssl_resnext101_32x8d	81.616	88.79	0.0455	21
267	gluon_resnext101_64x4d	80.604	83.46	0.0459	21
268	gluon_seresnext101_64x4d	80.894	88.23	0.0505	19
269	resnetrs200	84.066	93.21	0.0537	18
270	tf_efficientnet_b8	85.37	87.41	0.0551	18
271	tf_efficientnet_b8_ap	85.37	87.41	0.0552	18
272	hrnet_w30	78.206	37.71	0.0556	17
273	pnasnet5large	82.782	86.06	0.0556	17
274	hrnet_w18	76.758	21.3	0.0564	17
275	hrnet_w32	78.45	41.23	0.0572	17
276	nasnetalarge	82.62	88.75	0.0588	17
277	dpn131	79.822	79.25	0.0593	16
278	dm_nfnet_f1	84.604	132.63	0.0601	16
279	hrnet_w48	79.3	77.47	0.0609	16
280	hrnet_w40	78.92	57.56	0.061	16
281	hrnet_w44	78.896	67.06	0.0614	16
282	hrnet_w64	79.474	128.06	0.0629	15
283	dpn107	80.156	86.92	0.0634	15
284	ecaresnet269d	84.976	102.09	0.0635	15
285	regnety_320	80.812	145.05	0.0651	15
286	resnest200e	83.832	70.2	0.0651	15
287	regnetx_320	80.246	107.81	0.0657	15
288	resnetrs270	84.434	129.86	0.0722	13
289	gluon_senet154	81.234	115.09	0.0723	13
290	legacy_senet154	81.31	115.09	0.0727	13
291	ig_resnext101_32x16d	84.17	194.03	0.0784	12
292	ssl_resnext101_32x16d	81.844	194.03	0.0785	12
293	swsl_resnext101_32x16d	83.346	194.03	0.0785	12
294	resnetv2_50x3_bitm	83.784	217.32	0.0823	12
295	resnest269e	84.518	110.93	0.0866	11
296	dm_nfnet_f2	84.99	193.78	0.0873	11
297	resnetrs350	84.72	163.96	0.0931	10
298	resnetrs420	85.008	191.89	0.1106	9
299	resnetv2_152x2_bitm	84.44	236.34	0.1108	9
300	dm_nfnet_f3	85.56	254.92	0.1143	8
301	dm_nfnet_f4	85.658	316.07	0.1413	7
302	resnetv2_101x3_bitm	84.394	387.93	0.1557	6
303	ig_resnext101_32x32d	85.094	468.53	0.1678	5
304	dm_nfnet_f5	85.714	377.21	0.1686	5
305	tf_efficientnet_l2_ns_475	88.234	480.31	0.1875	5
306	tf_efficientnet_l2_ns	88.352	480.31	0.1877	5
307	dm_nfnet_f6	86.296	438.36	0.1959	5
308	ig_resnext101_32x48d	85.428	828.41	0.2854	3
309	resnetv2_152x4_bitm	84.932	936.53	0.3693	2